On Interface Issues in LES/RANS Coupling Strategies: A Method for Turbulence Forcing

Abstract

The present work deals with a computational strategy coupling near-wall, eddyviscosity-based RANS models with LES within a zonal Hybrid LES/RANS (HLR) framework. Key questions concerning the coupling of both methods, the inherently steady RANS method and highly-unsteady LES method, are closely connected to the treatment at the interface separating both sub-regions. Large attention was paid to this problem and following three issues were highlighted: (1) the exchange of the variables across the LES/RANS interface was adjusted by implicit imposition of the condition of equality of the modelled turbulent viscosities (by assuming the continuity of their resolved contributions across the interface), enabling a smooth transition from the near-wall RANS layer to the off-wall LES sub-region; (2) utilisation of a dynamic, flow-dependent interface position in the course of the simulation. The control parameter k * representing the ratio of the modelled (SGS) to the total turbulent kinetic energy in the LES region, averaged over all grid cells at the interface on the LES side, is adopted; (3) the third issue, the present work is focussing on, addresses the usage of a special forcing technique, which compensates the loss of information due to strong damping caused by the presence of the RANS region (the typical outcome of such a circumstance is the so-called velocity mismatch in the region of interface) by creation of artificial and correlated fluctuations using a method originating from a digital-filterbased generation of inflow data for spatially developing DNS and LES due to Klein et al. (2003). Herewith, the recovery of the fluctuations on the LES side of the interface is accelerated and the afore-mentioned velocity bump is eliminated to a largest extent. The performances of the model are illustrated against the available DNS and fine-grid LES of periodic flows in a plane channel and over a 2-D smoothly contoured hill respectively.

Numerical assessment of wall modelling approaches in scale-resolving in-cylinder simulations

“…Various interface positions between O(50)-O(600) have been applied with zonal approaches (e.g. Piomelli et al, 2003;Temmerman et al, 2005;Jakirlić et al, 2011). Based on numerical tests in the present setup (not shown herein for brevity), re-sults were not noted to be very sensitive to the exact position of the interface, corresponding with previous canonical flow observations (Keskinen et al, 2017).…”

“…In the HLR approach, Eqs. (1)-(4) incorporate an effective filter width (Jakirlić et al, 2011;Sagaut et al, 2013): in the RANS zone, the model mimics an SGS model whose length scale corresponds to the SGS width at the LES/RANS interface, approaching the scale l RAN S = k 3/2 /ε close to the wall (Jakirlić et al, 2011). As noted in our previous work (Keskinen et al, 2017)…”

“…Eq. 13 (Jakirlić et al, 2011) for a dynamic approach). The RANS zone employs a low-Reynolds k − ε turbulence model based on the model of Lien et al (1996) with pertinent modifications detailed in our previous study (Nuutinen et al, 2014).…”

“…(9)-(10)) in the first cell of the LES domain via source terms (cf. (Jakirlić et al, 2011) for a more detailed description). Interfaces in this study are placed manually to a scaled wall-normal distance of y + 1 ≈ O(100) in maximum gradient conditions, while intake jet regions were set in the LES zone.…”

Validation of a zonal hybrid URANS/LES turbulence modeling method for multi-cycle engine flow simulation

“…30 In a more general class of zonal approaches, where completely different models are used for the URANS and LES zones, a special interface treatment would have been unavoidable. 35,39,41,42 This, however, poses the additional question on how to correctly reconstruct URANS quantities from the resolved LES field and vice versa , often leading to sharp interface discontinuities and/or modeling issues. 39,42 One of our main objectives is to verify whether the relatively simple ZDES-type methodology, described by equations (7) and (8), could be effectively applied to actual engine geometries.…”